Connecting Hellos to AI: less admin pressure for social workers
Hellos can be connected to AI. IntraGPT talks to the Hellos modules, from Person and ContactPerson to Traject and TrajectMatch, plus the custom modules created for your organisation. Per module you set whether the agent may read, write or both. A social worker can therefore ask a question about a running trajectory without clicking through screens, and have a note prepared without filling in a form.
What the agent can do in Hellos
Explore the modules
The agent first asks which modules this connection holds and which of them allow reading, writing or both. That way the connector also works when your custom modules are named differently from another organisation's.
Search a module with intent
Searching happens with filters and with a small number of results per page, until it is clear which record is meant. That saves time and limits how much client data appears on screen.
Retrieve one record
Once the id is known the agent retrieves the full record. That is the moment it sees the details, not before.
Register and update
On the modules where that is allowed the agent can create or update a record, for example a contact moment or a progress note. It only does so when asked.
Link client and trajectory
TrajectMatch holds the relationship between a client and a running trajectory. Questions such as which trajectory is this person in and since when can be answered with it.
How community organisations use the connector
Social work
- Situation
- A social worker returns from a home visit and has to record the contact moment, but the admin keeps getting pushed back.
- What the agent does
- The worker dictates or briefly types what was discussed and the agent prepares it as a progress note in the right module against the right person.
- Result
- Registration happens the same day and the note reads like a note, not like scattered keywords.
Trajectory support
- Situation
- A support worker takes over a trajectory from a colleague and wants to know what has happened so far.
- What the agent does
- The agent finds the person, retrieves the trajectory match and summarises the contact moments from the relevant modules.
- Result
- A handover in a few minutes instead of an afternoon of reading files.
Team coordination
- Situation
- A coordinator wants to know which participants took part in a given activity this month.
- What the agent does
- The agent searches the custom module for that activity with a filter on the period.
- Result
- An overview without anyone having to build an export or a report.
Accountability
- Situation
- For the accountability report to the municipality, figures per trajectory have to be gathered.
- What the agent does
- The agent retrieves the counts per trajectory and lays them side by side, naming the modules the figures came from.
- Result
- A checkable first draft the policy officer can build on.
Setting it up carefully
-
1
Token and module catalogue
The connector works with a key you issue yourself and can revoke again. Which standard and custom modules exist follows from your own Hellos environment; the assistant sees no more modules than are released there.
-
2
A mode per module
For every module you pick read, write or both. A module you do not switch on does not exist for the agent. That keeps sensitive parts out of view.
-
3
Roles and agents
You decide which agents and which roles may use the connector. An agent for team coordination can see different modules than an agent for a support worker.
-
4
Watching the log
Every call lands in the connector log. If there is ever doubt about what the agent retrieved or wrote, it can be traced, without the content itself sitting in the log.
Special category data deserves extra restraint
Hellos holds data about vulnerable people. The connector is deliberately built so the agent fetches no more and says no more than the question requires.
Where the data lives: Nederland of de EU
Only what is needed
The agent searches with small result sets and only retrieves a full record once the id is certain. Whole case files are never copied into an answer.
Writing is off by default
A module can only be changed once you switch that on, and the agent writes only when explicitly asked.
Permissions follow the role
Who may use the agent is decided per role in IntraGPT. Not every employee reaches every module.
On your own server
IntraGPT runs on your own server with open-weight models. Client data is not sent to an external model provider and is not used to train a model.
Questions about the Hellos connector
Can I connect Hellos to ChatGPT?
Not responsibly. Hellos holds special category personal data, which does not belong in a public chatbot. IntraGPT offers the same way of working, but with a connector bounded per module, on your own server, with logging per call.
Does IntraGPT work with Hellos?
Yes. Hellos has a native implementation in IntraGPT with a module catalogue and actions to explore modules, search, retrieve a single record and create or update on the permitted modules.
Does it work with our own custom modules?
Yes, that is the whole premise. Alongside the standard modules the catalogue contains your organisation's custom modules, and you choose the mode per module yourself.
Can the AI change something in a client file?
Only on modules where you enabled writing, and only on explicit request. The agent shows what it wants to record before it does so, giving the support worker a chance to correct it.
What does a Hellos connector cost?
The connector is part of the project we run together. The work sits mainly in walking through the modules and deciding the permissions. In a thirty minute call we look at which modules solve the most.
How long does setup take?
With a token and the module catalogue the connection stands quickly. The time goes into the conversation about which modules may be included and who may use them; deliberately not a tick box exercise.
Is this GDPR compliant?
The connector is built so that no more is retrieved than needed, that writing has to be switched on explicitly and that every call is traceable. IntraGPT is ISO 27001 certified and runs on your own server. We walk through the processing agreement and the role setup together.
Can we combine this with our protocols?
Yes, and that is often the most useful combination. The connector supplies the facts from the file, the knowledge base supplies the protocol or work instruction. The agent states for each part of the answer where it came from.